Type alias TargetFn<T>

Type declaration

• • (location: T): number | Result

  • The interface that a function to be minimised must satisfy.

    • For Simplex, T should be number[] representing a point in n-dimensional space
    • For Brent, T should be number representing a single real value

    For both, the return value must either be a number (the objective value to be minimised) or Result (a rich result type that includes the objective value and additional information; see below).

    Remarks

    Sometimes, it is useful to get additional information out of the target function and store it alongside the solution - this is what the Result return type is for. For example, if fitting a model to data you might end up with a sum-of-squares error used for target fitting, but the model values themselves are of interest. For linear regression we might do this by returning a closure that could be evaluated at any x position:

    const xObserved = [0, 1, 2, 3, 4];
    const yObserved = [0.64, 2.41, 4.29, 6.62, 8.31]
    function target(theta) {
        const c = theta[0]; // intercept
        const m = theta[1]; // slope
        var tot = 0;
        for (var i = 0; i < xObserved.length; ++i) {
            tot += (c + m * xObserved[i] - yObserved[i])**2;
        }
        return {value: tot, data: (x) => x.map(el => c + m * el)};
    }
    

    This can be passed through to Simplex above, and the data element of Simplex.result() will contain the predictor function at the best point. (Do not fit linear regressions with this package - you can do that faster and more accurately with many other methods.)

    Returns

    result The value of the target function at location, either as a simple number or a rich Result type

    Parameters

    • location: T

      A point in n dimensional space to evaluate

    Returns number | Result